INDEX

Explanations

information related to press statements and efforts to reach a spokesperson for comments or responses

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 badass

-0.79

 dominating

-0.77

 conquering

-0.76

 effortlessly

-0.73

 smack

-0.73

 envy

-0.72

 brilliantly

-0.72

 advant

-0.72

 dominance

-0.71

 monopol

-0.71

POSITIVE LOGITS

 clarification

1.21

 pending

1.12

 confirm

1.06

 spokeswoman

1.06

 clarified

1.06

 confir

1.05

 redacted

1.04

 confirmed

1.04

confirmed

1.03

 requested

1.02

Activations Density 8.992%

information related to press statements and efforts to reach a spokesperson for comments or responses

No Comments

No Known Activations